Inefficient Communication
Today, more and more AI agents make phone calls. Also, more and more AI agents take phone calls.
Inevitably, AI agent would have to talk to AI agent over the phone.
Plain english wouldn't be perfect in this scenario:
- Inefficient time-wise
- Relatively high compute costs of voice generation
- Error prone
Instead, we propose AI agents to move from speech level to sound level protocol to communicate to each other.
The protocol
Gibber Link is a communication protocol, that handles following scenarious:
- If AI agent speaks to a human, it sticks to speech
- If AI agent detects another AI agent on the other side, it suggests to switch to sound level protocol
- If another AI agent accepts the protocol, they switch to sound level protocol (currently based on GGWave library)
Impact
- Gibber Link allows to avoid speech generation, which contribute to 90%+ of compute costs
- Gibber Link allows to communicate the same information in shorter (up to 80%) time
- Gibber Link is an error-proof protocol, allowing for clearer communication in noisy envinronments
- Gibber Link could be expanded to pass multi modal (images, ...) and structures (json, ...) data
Built With
- agent
- elevenlabs
- nextjs
- node.js
Log in or sign up for Devpost to join the conversation.